Problems with SOL 111 and superelements

Question

I created an SOL 111 analysis which runs well. After that I wanted to create superelements and make the same analysis, but I get an error:

* APPLIED LOADS

*** USER FATAL MESSAGE 3046 (SQFREQ)

THIS FREQUENCY RESPONSE ANALYSIS HAS NO EXCITATION SPECIFIED FOR IT.

APPLIED LOADS AND ENFORCED MOTION ARE BOTH NULL.

THE RESULT WILL THEREFORE BE A ZERO SOLUTION.

0FATAL ERROR

I defined the node, where the force attack, as a boundary point for the superelement. What is the reason that the analysis doesn't work?

If you need any file, I uploaded the files with the superelement analysis.

Attached Files (1)

Platte.zip

Don.Graff · Accepted Answer

Hi- it is always best if you can put any 'point load' type loading into the residual, as this greatly simplifies the SE setup, especially for dynamics. I made some changes to your input file:

1) I added modal DOF as placeholders for component modes for each SE... otherwise, no component modes get used in the assembly, which will greatly reduce the accuracy.

2) I removed the LOADSET/LSEQ setup... this is not necessary and can confuse things... it is OK to just point directly from the dynamic loading entry (RLOAD) to the applied load entry (FORCE)

3) I used Expanded Case Control... this allows you to cleanly define what is to be done for each SE... otherwise, it tries to do everything to all SE... for example, the SPC is only in SE 1, not in SE 2 or SE 0... so calling out SPC in the Subcase specifically for SE 1 is best (avoids warning messages).

4) I put Grid 11 from SE 2 into the ASET so it will be passed down to SE 0 for load application, and then copied the GRID definition to SE 0. I did this to be explicit in my desire to put this grid on the boundary... the automatic way to do it is to just make sure this grid also exists in the residual (SE 0) and it will automatically be put on the boundary... but I prefer to be explicit.

5) I renumbered a few QUAD's to eliminate duplicate element ids between the SE's so I could more easily create a baseline, non-SE version for checkout purposes. I suspect you didn't intend to duplicate these few elements in both SE1 and SE2... check elements 4, 5, 6, 14, 15, 24, 25, 34, 35. But I left this in my versions just in case you did intend that.

I think this is all I did, but will include my version of your input files for you to review.

Note that is is best to do your modal reduction of the SE to a prescribed upper cutoff frequency rather than to a specific number of modes... in this case it seems to work fine, but in general, you want to make sure each SE is solved to a frequency at least 2-5 times higher than the desired downstream assembled frequency. You can keep increasing this upstream cutoff frquency until you achieved the desired accuracy in comparison with the baseline run.

Regards,

Don

Attached Files (1)

Versuch1FRSuper_djg.zip